Qengineering / TensorFlow_Lite_SSD_RPi_32-bits

TensorFlow Lite SSD on a bare Raspberry Pi 4 at 17 FPS
https://qengineering.eu/install-tensorflow-2-lite-on-raspberry-pi-4.html
BSD 3-Clause "New" or "Revised" License
16 stars 5 forks source link

Customize training #2

Open adlerschilling opened 3 years ago

adlerschilling commented 3 years ago

I would like to customize a specific training, could you guide me which one should I use so that it works with the available code?

https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/tf1_detection_zoo.md#mobile-models

I've already created a training using ssd_mobilenet_v1_quantized_coco and ssd_mobilenet_v1_coco both did not work.

Qengineering commented 3 years ago

More people asked the same question. Up till now, we haven't found the time to write a decent guide on transfer training a TensorFlow SSD. Sorry. However, on the internet, there are several tutorials on the subject. Perhaps this one helps you to start custom detection.

adlerschilling commented 3 years ago

I just wanted to know which one you used to generate the detect.tflite that you have in your project?

Qengineering commented 3 years ago

I have used the models from the official site without bothering to convert them. TensorFlow Lite models